in the email package (2.4.1) the get_filename() method
returns the MIME field "filename" but some messages
have 'name' field instead, for example:

USUALLY THE HEADER IS:
Content-Type: application/octet-stream;
        name="XX.pdf"
Content-Transfer-Encoding: base64
Content-Description: XX.pdf
Content-Disposition: attachment;
        filename="XX.pdf"

BUT SOMETIMES THE HEADER IS:
Content-type: application/octet-stream; name="XX.xls"
Content-transfer-encoding: base64

For this to work properly I had to code a hack along
these lines:
filename = part.get_filename()
if not filename:
   ct = part.get("Content-type")
   m = re.compile('name=\"(\S+)\"').search(ct, 1)
   if m: filename=m.group(1)

But it would be helpful to code this in the get_filename()

Michal

Logged In: YES 
user_id=12800

r42075 for Python 2.3 / email 2.5 (will close after porting
to all other branches).

Logged In: YES 
user_id=12800

r42076 for email 3.0 / python 2.5

r42077 for python 2.4

Looks like this one needs reopening to me... I've recently had to parse out attachments with the following Content-Type lines and no Content-Disposition provided:

Content-Type: application/vnd.ms-excel; name=transactions.xls

It might not seem like much, but the content-type check for this workaround in get_filename appears to be case-sensitive, and thus misses the upper case "T" in Content-Type.
There is no information provided in the headers about the mailer they use, but they get plenty of other bits wrong too (eg. the xls file is actually just tab-separated values).

A slight update for my workaround for the above with the following code:

if not filename:
   ct = part.get("Content-Type")
   if ct:
      m = re.compile('name=\"?(\S+)\"?').search(ct, 1)
      if m: filename = m.group(1)

I've added ? operators to the double-quotes, and changed the case in the part.get. However, there may be good reasons as to why part.get needs to be case-sensitive. Not my area of expertise though.

Ich, if your problem still exists in 2.7, 3.1, or 3.2, please open a new issue with a test case showing the problem you are running in to.

Also, issue 7082 might be relevant here, since it fixed a bug in this fix.